新的集成预报及其在短期气候预测中的应用

分析了传统的基于加权的集成预报等方法及其在气象预测应用中存在的问题,在此基础上提出了一种新的基于数据挖掘的集成预报方法,并选用BP人工神经网络建立集成预报分类器来对各种子预报方法的预报结果进行集成和综合;该方法可以根据不同预报对象的特性,对集成预报权值进行动态改变,克服了传统的集成预报方法中权值一旦确定就不能改变的不足,也克服了现有的集成预报不能得到最优结果的不足。通过对2001～2007年重庆市城口县1月的降水和平均气温以及重庆市的春旱指数进行预报,实验结果显示,集成预报结果的可靠性和准确性不但高于集成之前的各种子预报方法,而且高于传统的其它集成预报方法,验证了方法的有效性。

A new integrating forecast and its application in short-term climate prediction

The existing problems of the traditional weight integrating forecast methods and the application in climate prediction are analyzed. A new method based on data mining is presented, which uses BP artificial neural network to build the integrating forecast classifier to integrate the forecast results of sub-methods. According to the features of different forecast objects, this method can change weight dynamically, which overcomes the shortage of the traditional weight integrating forecasts that cannot change weight after been decided and overcomes the shortage that cannot get the optimal results. By predicting the precipitation and average temperature of Chengkou County in January, and spring drought index of Chongqing from 2001 to 2007, the experiment results show that the reliability and accuracy of the proposed model are better than those of the sub-methods and other integrating forecast methods, which proves the effectiveness of this method.